Search

In financial field, a robust IT system is of vital importance to ensure the smooth operation of financial transactions. However, many financial corporations still depend on operators to identify and eliminate the system failures when financial IT systems break down. This traditional operation method is time consuming and extremely inefficient. To improve the efficiency and accuracy of system failure detection and thereby reduce the impact of system failures on financial services, we propose a novel machine learning-based framework to predict the occurrence of system exceptions and failures in a financial IT system. In particular, we first extract rich information from system logs and eliminate noises in the data. Then the cleaned data is leveraged as the input of our proposed anomaly prediction framework which consists of three modules: key performance indicator data prediction module, anomaly identification module and severity classification module. Notably, we design a hierarchical architecture of alarm classifiers and try to alleviate the influence of class-imbalance problem on the overall performance. Empirically, the experimental results demonstrate the superior performance of our proposed method on a real-world financial IT system log data set. [ABSTRACT FROM AUTHOR]

In this paper, a novel decision tree classifier based on shallow neural networks with piecewise and nonlinear transformation activation functions are presented. A shallow neural network is recursively employed into linear and non-linear multivariate binary decision tree methods which generates splitting nodes and classifier nodes. Firstly, a linear multivariate binary decision tree with a shallow neural network is proposed which employs a rectified linear unit function. Secondly, there is presented a new activation function with non-linear property which has good generalization ability in learning process of neural networks. The presented method shows high generalization ability for linear and non-linear multivariate binary decision tree models which are called a Neural Network Decision Tree (NNDT). The proposed models with high generalization ability ensure the classification accuracy and performance. A novel split criterion of generating the nodes which focuses more on majority objects of classes on the current node is presented and employed in the new NNDT models. Furthermore, a shallow neural network based NNDT models are converted into a hyperplane based linear and non-linear multivariate decision trees which has high speed in the processing classification decisions. Numerical experiments on publicly available datasets have showed that the presented NNDT methods outperform the existing decision tree algorithms and other classifier methods.

Object location in indoor environments is challenging when there is no physical contact, a field of view, reflective materials, and an excess of obstacles. Several works propose using Radio Frequency Identification technology (RFID) and machine learning methods to develop location systems in those situations. However, using an object as a target class slows learning and prediction in large-scale environments. To circumvent such problems, we proposed a location system that uses hierarchical classification. We divided the environment into regions to reduce the classifier’s training and the number of predicted classes. To define the regions, we used clustering techniques, indicating which clustering technique achieves the best performance in the proposed scenario. The main contribution of this work is a high-precision location system for large-scale environments. The results showed the proposed system’s implantation in a real environment with 400 target objects with 5 cm of location precision. The accuracy for region detection is 99.36%, while for identifying the object is 99.94%. Additionally, with the proposed hierarchical approach, we showed a reduction of 38.16% and 58.39% in processing time and classifier model size. [ABSTRACT FROM AUTHOR]

Classifying point cloud of urban landscapes plays essential roles in many urban applications. However, automating such a task is challenging due to irregular point distribution and complex urban scenes. Incorporating contextual information is crucial in improving classification accuracy of point clouds. In this article, we propose a hierarchical approach for point cloud classification with 3-D contextual features, which comprises three steps:segment-based classification with primitive features and a random forest classifier; extracting novel 3-D contextual features from the initial labels considering spatial relationships between neighboring segments and semantic dependencies; and refining classification with a combination of primitive features and spatial contextual features, and a hierarchical multilayer perceptron classifier that considers primitive features and spatial contextual features at different levels. The proposed method was tested on two point cloud datasets:the National University of Singapore (NUS) dataset and the Vaihingen benchmark dataset of the International Society of Photogrammetry and Remote Sensing. The evaluation results showed that the proposed method achieved an overall accuracy of 92.51% and 82.34% for the NUS dataset and Vaihingen dataset, respectively. The feature importance evaluation showed that 3-D spatial contextual features contributed useful information for discriminating different classes, such as roof, facade, grassland, tree, and ground. Quantitative comparisons further showed that the proposed method is more advantageous, especially in the detection of class roof and facade.

This paper proposes the architecture of a convolutional neural network that creates a neural network system for recognizing objects in images using our own approach to classification using a hierarchical classifier. The architecture will be assigned to find the optimal solution to the problem for many sets of image data and, unlike existing approaches, will have high performance indicators without losing the number of parameters during recognition, and most importantly, the best value of object recognition accuracy compared to existing models of convolutional neural networks. The main attention is paid to the approach to training such a network and conducting experiments on the generated samples of various datasets using graphic processing units (GPUs). [ABSTRACT FROM AUTHOR]

A gestural input to control artifacts and access the digital world is an essential part of highly usable systems. In this article, we propose a gesture recognition method that leverages the sound generated by the friction between a surface such as a table and a finger or pen, in which 17 different gestures are defined. The gesture recognition process is regarded as a 17-class classification problem; 89 classification features are defined to represent the envelope of each input sound, while a hierarchical classifier structure is employed to increase the accuracy of confusable classes. Offline experiments show that the highest accuracy is 0.954 under a condition where the classifiers are customized for each user, while an accuracy of 0.854 is obtained under a condition where the classifiers are trained without using the data from test users. We also confirm the effectiveness of the hierarchical classifier approach compared with a single-flat-classifier approach and that of a feature engineering approach compared with a feature learning approach. The information of individual features is also presented. [ABSTRACT FROM AUTHOR]

• A hierarchical classifier is proposed to pre-impact fall detection. • Trained classifier model can be ported to embedded systems. • The multi-sensor-based method achieves a better sensitivity and specificity. • This study has potential application in a fall-injury prevention system. Fall is a major threat to elder health, and fall detection has attracted considerable research attention recently. In our study, a novel method is proposed to detect falls prior to impact during walking. Angle and angular-velocity data from the waist and thigh are collected using two wearable sensors. By extracting and selecting distinctive features, we aim to identify falls at an early stage. To improve detection accuracy and reduce false alarms, a hierarchical classifier based on Fisher discriminant analysis is developed. With the hierarchical classifier, human activities are classified into three categories: non-fall, backward fall and forward fall. It can achieve average lead times of 376 ms for backward fall and 404 ms for forward fall. Meanwhile, it can achieve a sensitivity of 95.5% and specificity of 97.3%. The method can achieve a high accuracy for classifier, and a long lead time for pre-impact fall detection. Compared with single-sensor-based methods, the multi-sensor-based method achieves a better performance. The preliminary results indicate that our study has potential application in a fall-injury prevention system. [ABSTRACT FROM AUTHOR]

Highlights • Extreme gradient boosting is first introduced and utilized for single heartbeat classification. • A weighted XGBoost classifier for unbalanced heartbeat dataset with multiple classes is presented. • A hierarchical classifier based on weighted extreme gradient boosting and threshold classifiers is constructed. • Recursive feature elimination is employed for feature selection from a large number of features. • Both high positive predictive value for N class and high sensitivity for abnormal classes are provided, which is practical for clinical diagnosis. Abstract Background and objective Electrocardiogram (ECG) is a useful tool for detecting heart disease. Automated ECG diagnosis allows for heart monitoring on small devices, especially on wearable devices. In order to recognize arrhythmias automatically, accurate classification method for electrocardiogram (ECG) heartbeats was studied in this paper. Methods Based on weighted extreme gradient boosting (XGBoost), a hierarchical classification method is proposed. A large number of features from 6 categories are extracted from the preprocessed heartbeats. Then recursive feature elimination is used for selecting features. Afterwards, a hierarchical classifier is constructed in classification stage. The hierarchical classifier is composed of threshold and XGBoost classifiers. And the XGBoost classifiers are improved with weights. Results The method was applied to an inter-patient experiment conforming AAMI standard. The obtained sensitivities for normal (N), supraventricular (S), ventricular (V), fusion (F), and Unknown beats (Q) were 92.1%, 91.7%, 95.1%, and 61.6%. Positive predictive values of 99.5%, 46.2%, 88.1%, and 15.2% were also provided for the four classes. Conclusions XGBoost was improved and firstly introduced in single heartbeat classification. A comparison showed the effectiveness of the novel method. The method was more suitable for clinical application as both high positive predictive value for N class and high sensitivities for abnormal classes were provided. [ABSTRACT FROM AUTHOR]

With their highly social nature and complex vocal communication system, marmosets are important models for comparative studies of vocal communication and, eventually, language evolution. However, our knowledge about marmoset vocalizations predominantly originates from playback studies or vocal interactions between dyads, and there is a need to move towards studying group-level communication dynamics. Efficient source identification from marmoset vocalizations is essential for this challenge, and machine learning algorithms (MLAs) can aid it. Here we built a pipeline capable of plentiful feature extraction, meaningful feature selection, and supervised classification of vocalizations of up to 18 marmosets. We optimized the classifier by building a hierarchical MLA that first learned to determine the sex of the source, narrowed down the possible source individuals based on their sex and then determined the source identity. We were able to correctly identify the source individual with high precisions (87.21%-94.42%, depending on call type, and up to 97.79% after the removal of twins from the dataset). We also examine the robustness of identification across varying sample sizes. Our pipeline is a promising tool not only for source identification from marmoset vocalizations but also for analysing vocalizations of other species.

With smartphones increasingly becoming ubiquitous and being equipped with various sensors, nowadays, there is a trend towards implementing HAR (Human Activity Recognition) algorithms and applications on smartphones, including health monitoring, self-managing system and fitness tracking. However, one of the main issues of the existing HAR schemes is that the classification accuracy is relatively low, and in order to improve the accuracy, high computation overhead is needed. In this paper, an efficient Group-based Context-aware classification method for human activity recognition on smartphones, GCHAR is proposed, which exploits hierarchical group-based scheme to improve the classification efficiency, and reduces the classification error through context awareness rather than the intensive computation. Specifically, GCHAR designs the two-level hierarchical classification structure, i.e., inter-group and inner-group, and utilizes the previous state and transition logic (so-called context awareness) to detect the transitions among activity groups. In comparison with other popular classifiers such as RandomTree, Bagging, J48, BayesNet, KNN and Decision Table, thorough experiments on the realistic dataset (UCI HAR repository) demonstrate that GCHAR achieves the best classification accuracy, reaching 94.1636%, and time consumption in training stage of GCHAR is four times shorter than the simple Decision Table and is decreased by 72.21% in classification stage in comparison with BayesNet. [ABSTRACT FROM AUTHOR]

In this paper, a new hierarchical method of contour analysis is proposed. The first stage of the analysis is based on the syntactic approach in which elementary segments of the contour are extracted by using geometric features. As a result, each extracted segment belongs to one of the equivalence classes which make the set of primitives. The contour is described as a string of primitives. The applied syntactic approach is a generalization of a classical shape language. Such description enables to classify some cases, but not all. Therefore, the second stage of the classification is applied. It consists in analysing other local geometric features of the contour in the context of the syntactic description obtained at the first stage of the analysis. The values of these features are used to define the arguments of the membership functions of the family of fuzzy sets which is the basis for the contour classification at the second stage of the analysis. The proposed formalism has been applied effectively to the recognition of bone contours in X-ray images. Healthy bones, the bones with erosions and the bones with osteophytes have been classified correctly. The proposal described in this paper turned out to be effective in the classification of contours of bones. [ABSTRACT FROM AUTHOR]

Face is one of the important parts of the human body that can be used in video surveillance security (VSS) system for identity recognition purposes. However, systems that work under uncontrolled environment such as VSS system suffer from illumination changes, unpredictability of face appearance due to the presence of accessories such as sunglasses and scarf, connected face and multiple face sizes. In this paper, a novel algorithm known as Hierarchical Skin-AdaBoost-Neural Network (H-SKANN) is introduced to overcome these problems. Skin is used to roughly locate face candidates. Then, AdaBoost is used to filter out non-face candidates. Subsequently, an artificial neural network is utilized as the main filter to finally detect the face. In order to handle multiple face sizes, all these algorithms are arranged in hierarchical manner. On top of this, face skin merging (FSM) is also introduced to connect blobs of skin regions to form a face. Experiments conducted on six single-face databases (AR, FERET, IMM, Georgia, Caltech, and Talking-PIE) and one multi-face benchmark database (ChokePoint) demonstrated that 98.07% and 95.48% of averaged accuracy have been achieved for single- and multi-face detection, respectively, using the proposed method. [ABSTRACT FROM AUTHOR]

In the present work, a hybrid hierarchical framework for classification of breast density using digitized film screen mammograms has been proposed. For designing of an efficient classification framework 480 MLO view digitized screen film mammographic images are taken from DDSM dataset. The ROIs of fixed size i.e. 128 × 128 pixels are cropped from the center area of the breast (i.e. the area where glandular ducts are prominent). A total of 292 texture features based on statistical methods, signal processing based methods and transform domain based methods are computed for each ROI. The computed feature vector is subjected to PCA for dimensionality reduction. The reduced feature space is fed to the classification module. In this work 4-class breast density classification has been conducted using hierarchical framework where the first classifier is used to classify an unknown test ROI into B-I/ other class. If the test ROI is predicted as other class, it is inputted to second classifier for the classification into B-II/ dense class. If the test ROI is predicted as belonging to dense class, it is inputted to classifier for the classification into B-III/ B-IV class. In this work five hierarchical classifiers designs consisting of 3 PCA- kNN, 3 PCA-PNN, 3 PCA-ANN, 3 PCA-NFC and 3 PCA-SVM classifiers has been proposed. The obtained maximum OCA value is 80.4% using PCA-NFC in hierarchical approach. Further, the best performing individual classifiers are clubbed together in a hierarchical framework to design hybrid hierarchical framework for classification of breast density using digitized screen film mammograms. The proposed hybrid hierarchical framework yields the OCA value of 84.1%. The result achieved by the proposed hybrid hierarchical framework is quite promising and can be used in clinical environment for differentiation between different breast density patterns. [ABSTRACT FROM AUTHOR]

Emotion recognition in the wild is a very challenging task. In this paper, we investigate a variety of different multimodal features (acoustic and visual) from video clips to evaluate their discriminative abilities in human emotion analysis. For each clip, we extract MSDF BoW, LBP-TOP, PHOG, LPQ-TOP and Audio features. We train different classifiers for every type of feature on the AFEW dataset from the ICMI 2014 EmotiW Challenge, and we propose a novel hierarchical classification framework, which combines the feature-level and decision-level fusion strategy for all of the extracted multimodal features. The final achievement we gain on the AFEW test set is 47.17 %, which is considerably better than the best baseline recognition rate of 33.7 %. Among all of the teams participating in the ICMI 2014 EmotiW challenge, our recognition performance won the first runner-up award. Furthermore, we test our method on FERA and CK datasets, the experimental results also show good performance. [ABSTRACT FROM AUTHOR]

In the binary space of large dimension we analyze the nearest neighbor search problem where the required point is a distorted version of one of the patterns. Previously it was shown that the only algorithms able to solve the set problem are the exhaustive search and the neural network search tree. For the given problem the speed of operation of the last algorithm is dozens of times larger comparing with the exhaustive search. Moreover, in the case of large dimensions the neural network tree can be regarded as an accurate algorithm since the probability of its error is so small that cannot be measured. In the present publication, we propose a modification of the scalar neural network tree allowing the speeding of the algorithm's operation up to hundred times without losses in its reliability. [ABSTRACT FROM AUTHOR]

Human action recognition from video has several potential to apply in different real-life applications, but the most cases in this field suffer from the variation in viewpoint. Most of published methods in this area are considered the performance of each single camera, therefore the change in the viewpoints significantly decrease the recognition rate. In this paper, multiple views are considered together and a method has proposed to recognize human action depicted in multi-view image sequences. In the first step, the border of the human body's silhouette is extracted and distance signal is calculated. In the next step, the wavelet transform is applied to extract coefficients of single-view features, and then the extracted features are combined to compose multi-view features. Finally a hierarchical classifier using support vector machine and Naïve Bayes classifiers is implemented to classify the actions. The average of overall action recognition accuracy for 12 actions using 5 different angles of views on the IXMAS dataset is 88.22. The results of experiments on the popular multi-view dataset have shown the proposed method achieves high and state-of-the-art success rates. In other word, combination of single-view extracted features from the wavelet approximation coefficients and composing the multi-view features can be used as the multi-view features. Further, the hierarchical classifier can be applied to recognize actions in multi-view human action recognition area. [ABSTRACT FROM AUTHOR]

Territorial planning and management requires that the spatial structure of the socioecological sectors is adequately understood. Several classification techniques exist that have been applied to detect ecological, or socioeconomic sectors, but not simultaneously in the same model; and also, with a limited number of variables. We have developed and applied a new probabilistic methodology – based on hierarchical hybrid Bayesian network classifiers – to identify the different socioecological sectors in Andalusia, a region in southern Spain, and incorporate a scenario of change. Results show that a priori , the socioecological structure is highly heterogeneous, with an altitude gradient from the river basin to the mountain peaks. However, under a scenario of global environmental change this heterogeneity is lost, making the territory more vulnerable to any alteration or disturbance. The methodology applied allows dealing with complex problems, containing a large number of variables, by splitting them into several sub-problems that can be easily solved. In the case of territorial planning, each component of the territory is modelled independently before combining them into a general classifier model. Furthermore, it can be applied to any complex unsupervised classification problem with no modification to the methodology. [ABSTRACT FROM AUTHOR]